Abstract:
Currently, climate change, floods, and other extreme weather events are becoming increasingly common, resulting in significant impacts on society. Predicting streamflow is essential for effective hydrology and water resource management. Accurate and timely predictions of streamflow help in water allocation, flood forecasting, and reservoir management. Forecasting strearnflow in cold regions presents chaHenges due to the significant annual and the seasonal changes in the natural processes occurring within catchments. The research aims to contribute to the understanding of stream flow dynamics in scarcely gauged catchments and to offer significant perspectives regarding water resource management in the study area. The purpose of this study is to evaluate the application of Artificial-Neural-Network (ANN) and Snowmelt run off model for streamflow prediction over scarcely gauge catchments like Astore river, Hunza and Gilgit river. For this purpose, Climatic variable data, including precipitation, temperature and Observed Streamflow data are obtained from reputable sources like MeteorologicalDepartment and Water & Power Development Authority in Pakistan. Snow cover data from MODIS imageries and DEM from USGS earth explorer. Three different ANNs models are implemented, The Keras Sequential Feedforward Single Layer Perceptron (SLP) Model, The Keras Sequential Feedforward with Backpropagation Model MultiLayer-Perceptron and Radial-Basis-Function Models are used. In results Multi-Layer Perceptron (MLP) model performed effectively and yielded the best results RA2 98% and NSE 99% during the validation and testing phases as compare to others. The findings of this research indicate the effectiveness of artificial neural network models in capturing the complex relationships between climatic variables, snow cover/glacier dynamics, and stream flow variability, achieving high predictive accuracy across different temporal scales and spatial locations within the study area as compare to Snowmelt runoff model.
